Mobile SLAM mapping platforms like Nexys will often reference point clouds as the end result of their scanning, but can sometimes gloss over the critical step of point cloud cleaning. Point cloud cleaning is vital to improve the quality, shareability, and ease with which the point cloud can be used. Point cloud cleaning also allows the scanning process to be far more forgiving since you can clean up noise like ghosting afterwards. This further simplifies the capture process, allowing even faster surveying in a variety of challenging scenarios.
Below, we’ll cover what you need to know about point cloud cleaning and what options are available via the ExynView post-processing suite.
A point cloud is the foundation of an environmental 3D digital twin. When a mapping unit like our Nexys moves through a space, it emits laser light. Each reflection of that laser light from a surface in the environment creates a single “point”. Each point contains position data in three-dimensional space, an 
intensity value and an RBG value if the Nexys cameras are active.
However, not all of those points are always needed for every application, and this is where point cloud cleaning enters the equation. Point cloud cleaning is the process of removing or altering points to improve the point cloud or shrink its overall file size without degrading accuracy.
Raw point cloud data can be a lot to process into an accurate 3D model. LiDAR sensors collect large volumes of information that is validated by ExynAI at thousands of times per second to create an accurate 3D model after post processing. This approach makes mobile surveying more flexible, but will require an extra cleaning step.
Point cloud cleaning helps eliminate any unwanted points to produce a more compact, accurate, and application-ready dataset. This is especially critical when point clouds need to be:
A cleaner point cloud is faster to render, easier to manipulate, and more widely compatible (not to mention better looking), making it far more useful across a range of tasks. The real art of point cloud cleaning is finding a perfect balance between removing unnecessary data while preserving the quality and resolution required for the surveying task.
Point cloud cleaning generally falls under three different objectives. Those three objectives are subsampling, point cloud smoothing, and the removal of non-static objects. Each of these processes can use different methods for achieving optimal results, depending on the application and requirements of the final point cloud.
The main goal of subsampling is to reduce the file size of the point cloud while maintaining the necessary point cloud density for the specific application. You can think of subsampling as similar to image or video compression techniques that are used for media streaming services like Netflix. In point clouds this is accomplished through advanced algorithms examining a collection of points and keeping 10 points where there used to be 100. The goal is to reduce the file size by removing information that the user will not notice is missing. The end result is a much smaller file size, easier shareability, and faster rendering or processing.
Non-static objects within a point cloud are any objects that were moving during the point cloud capture. This can include workers on the survey site, vehicles like a truck passing through, or any object that only temporarily entered the survey area and is not necessary for the final point cloud.
Non-static object removal is beneficial because it allows mobile scanning to be performed without the need to shut down a job site or survey area. This form of point cloud cleaning also allows you to survey in more challenging environments.
Screenshots of a point cloud before and after non-static point removal
Point cloud smoothing is used to remove small surface irregularities in geometric structures. These irregularities can be caused by noise or slight variations when the laser light is reflected back to the sensors.
If you’re familiar with photo retouching tools such as Photoshop, you can think of point cloud smoothing as similar to anti-aliasing, which smooths digital artifacts that can cause lines to appear jagged. Point cloud smoothing does something similar for flat three-dimensional surfaces, such as walls and floors. It can also be used when the fine detail of a surface is not needed.
For example, when measuring a mine cavity, the exact detailed texture of the walls is not needed for volumetric calculations, so these features can be removed by point cloud smoothing. Point cloud smoothing can also be useful for architectural visualizations meant to mimic the final structure’s smooth and flat appearance.
Point cloud smoothing can also be used to enhance the clarity of the point cloud. For example, when scanning a historical building facade, smoothing can be applied to remove irregularities, and the result is a much more accurate 3D representation of the ornate structure.
You can think of point cloud cleaning like your surveying safety net. Even if conditions are not ideal during your scan, you can still capture a highly accurate point cloud and use point cloud cleaning to remove unnecessary details to drastically reduce file size and improve the quality.
However, not all 3D mapping systems have these features built directly into the hardware for immediate on-site processing. With Nexys, the included ruggedized tablet allows you to view the point cloud through our ExynView app. Within the app, you can clean the point cloud as needed to reduce the file size or remove unnecessary objects.
This unique feature within the Nexys ecosystem allows you to confirm you have the point cloud you need before leaving the surveying site. ExynView gives you the confidence to survey in more challenging environments, knowing that you can check and clean the point cloud on-site immediately after it’s captured.
If you want to learn more about point clouds and how Nexys delivers on-site point cloud cleaning in the palm of your hand, contact us today for a free live demo of our entire Nexys mobile mapping ecosystem.
